Development of a Miniaturized Mechanoacoustic Sensor for Continuous, Objective Cough Detection, Characterization and Physiologic Monitoring in Children With Cystic Fibrosis

Abstract

Cough is an important symptom in children with acute and chronic respiratory disease. Daily cough is common in Cystic Fibrosis (CF) and increased cough is a symptom of pulmonary exacerbation. To date, cough assessment is primarily subjective in clinical practice and research. Attempts to develop objective, automatic cough counting tools have faced reliability issues in noisy environments and practical barriers limiting long-term use. This single-center pilot study evaluated usability, acceptability and performance of a mechanoacoustic sensor (MAS), previously used for cough classification in adults, in 36 children with CF over brief and multi-day periods in four cohorts. Children whose health was at baseline and who had symptoms of pulmonary exacerbation were included. We trained, validated, and deployed custom deep learning algorithms for accurate cough detection and classification from other vocalization or artifacts with an overall area under the receiver-operator characteristic curve (AUROC) of 0.96 and average precision (AP) of 0.93. Child and parent feedback led to a redesign of the MAS towards a smaller, more discreet device acceptable for daily use in children. Additional improvements optimized power efficiency and data management. The MAS's ability to objectively measure cough and other physiologic signals across clinic, hospital, and home settings is demonstrated, particularly aided by an AUROC of 0.97 and AP of 0.96 for motion artifact rejection. Examples of cough frequency and physiologic parameter correlations with participant-reported outcomes and clinical measurements for individual patients are presented. The MAS is a promising tool in objective longitudinal evaluation of cough in children with CF.

Fig. 1.

A schematic view of key sensor components and data flow. The device stores raw data locally on onboard memory for low-power operation or streams to a mobile device through BLE. The mobile device is able to display data and run pre-defined signal processing locally. All raw data are securely uploaded to cloud storage from which retrospective analyses are performed and larger models can be trained.

Fig. 2.

Example segment of raw, surface-normal acceleration data during a coughing fit. The raw acceleration data (top panel) is converted into a Fourier power spectrogram (middle panel), from which “burst” events are detected via power-sum peaks of spectrally wide-band data. These events are further converted into wavelet scalograms (bottom panel), which better represent dynamic frequency changes and can simultaneously capture relationships between low frequency components (throat and chest wall motion) which are present in cough events (purple arrows) and not in simple vocalizations or throat clearing (dotted circle).

Fig. 3.

Overview of the CNN-based Cough Classifier Model

Fig. 4.

Recruitment flow diagram

Fig. 5.

New Device Schematic, a top-down circuit view with key components highlighted, b fully encapsulated side view with annotations.

Fig. 6.

Mechanoacoustic Cough Sensor, a, exploded view of the pediatric optimized device, b, a pair of fully encapsulated devices, c-d, multi-colored devices mounted on the suprasternal notch. Note the subjects must lower their shirt collar to fully display the device.

Fig. 7.

Raw Data and Spectrograms in Free-living Environments. Raw triaxial accelerometer data are shown above the corresponding Fourier power spectrogram with colors normalized for clarity, red corresponds to higher power, blue to lower power. (a) shows a typical morning routine for a CF patient while wearing the device. Examples of events captured, ranging from subtle oscillations of heart valves opening/closing and chest wall motion (b), CF-related mechanical vest therapy and superimposed coughing (c), and physical activity (d) are shown.

Fig. 8.

Cough Classification Testing Set Results. a, confusion Matrix, b, receiver-operator characteristic curves and c, precision-recall curves

Fig. 9.

Multi-day Continuous Vital Signs and Cough Tracking. A patient’s clinical course is continuously monitored from admission to discharge and resumption of daily activities. Average daily HR, RR, skin temperature, cough frequency and intensity decrease as survey-based symptom assessments improve, mechanical vest therapy is spaced, and the patient is clinically cleared to discharge home.